those detected in complete 30 year missing 20% data are ... · s e a wa t e r t e m p e ra t u re t...
1
Detecting marine heatwaves with sub- optimal data Robert W. Schlegel 1,2,* @wiederweiter [email protected] Eric C. J. Oliver 1 Alistair J. Hobday 3 Albertus J. Smit 2 1 Department of Oceanography, Dalhousie University, Halifax, Nova Scotia, Canada 2 Department of Biodiversity and Conservation Biology, University of the Western Cape, Bellville, South Africa 3 CSIRO Oceans and Atmosphere, Hobart, Tasmania, 7000, Australia 1 Introduction The Hobday et al. (2016) and Hobday et al. (2018) definitions are used here for marine heatwaves (MHWs) and their categories. These events are increasing in duration and intensity around the globe (Oliver et al. 2018). Many coastal, oceanic, and polar regions with ecosystems/species that are vulnerable to these events (Smale et al. 2019) do not have optimal sea water temperature time series available for the detection of MHWs. An optimal time series is defined here as meeting these two primary requirements: 1) 30 years long, 2) no missing data. Any time series shorter than 30 years long or missing any data is therefore sub- optimal. The purpose of this study was to determine how confident we may be in our results when using sub-optimal data. 1.1 Objectives 1. Find the sub-optimal data limits for detecting MHWs 2. Quantify the eects of sub-optimal data on individual MHWs 3. Determine if these eects are consistent around the globe 2 Methods Reference time series from the heatwaveR package (Schlegel and Smit 2018) were re-sampled 100 times and made increasingly sub-optimal by: Removing one year of data; 37 – 10 years Randomly removing 1% of data; 0 – 50% Adding linear trend of 0.01°C/dec.; 0.00 – 0.50°C/dec. The eect on an entire time series and one focus MHW (Figure 2.1) were measured The sub-optimal treatments were performed on global NOAA OISST product and eects measured Figure 2.1: Focus MHWs from the three reference time series: A) Western Australia, B) Northwest Atlantic, C) Mediterranean. Start/end dates shown with light-green marks. 3 Results MHWs detected with 10 years of data do not dier significantly from those detected with 30 years of data (Figure 3.1) 20% missing data or less does not have an appreciable eect on the MHWs in a time series, but this can impact individual MHWs by dividing them up into smaller events 3.2) Individual MHWs in 10 year time series around the globe are on average 10% less intense (Figure 3.3) and 28% shorter than the same event in a 30 year time series Figure 3.1: The eect of length on the MHWs detected within the 100 re-sampled time series. Note that lengths greater than 30 years also produce dierent results. Figure 3.2: The eect on the focus MHWs (Figure 2.1) given the three dierent sub-optimal data challenges. Figure 3.3: Global map showing the trends in increasing or decreasing maximum intensity of MHWs as a time series is lengthened. 4 Conclusions The MHW detection algorithm provides robust results that are insensitive to sub-optimal data challenges. Missing data has a greater eect on MHWs than time series length. Linear interpolation is an eective fix for missing data up to 40%. Time series length has a less predictable eect than missing data. The shifting up or down of the mean climatology is the primary driver of change in MHWs. Optimal data should still be used when possible. References Hobday, Alistair J, Lisa V Alexander, Sarah E Perkins, Dan A Smale, Sandra C Straub, Eric CJ Oliver, Jessica A Benthuysen, et al. 2016. “A Hierarchical Approach to Defining Marine Heatwaves.” Progress in Oceanography 141. Elsevier: 227–38. Hobday, Alistair J, Eric CJ Oliver, Alex Sen Gupta, Jessica A Benthuysen, Michael T Burrows, Markus G Donat, Neil J Holbrook, et al. 2018. “Categorizing and Naming Marine Heatwaves.” Oceanography 31 (2). JSTOR: 162–73. Oliver, Eric CJ, Markus G Donat, Michael T Burrows, Pippa J Moore, Dan A Smale, Lisa V Alexander, Jessica A Benthuysen, et al. 2018. “Longer and More Frequent Marine Heatwaves over the Past Century.” Nature Communications 9 (1). Nature Publishing Group: 1324. Schlegel, Robert W, and Albertus J Smit. 2018. “HeatwaveR: A Central Algorithm for the Detection of Heatwaves and Cold-Spells.” The Journal of Open Source Software 3 (27): 821. Smale, Dan A, Thomas Wernberg, Eric CJ Oliver, Mads Thomsen, Ben P Harvey, Sandra C Straub, Michael T Burrows, et al. 2019. “Marine Heatwaves Threaten Global Biodiversity and the Provision of Ecosystem Services.” Nature Climate Change 9. Nature Publishing Group: 306–12. Marine heatwaves detected in time series as short as 10 years and missing 20% data are comparable to those detected in complete 30 year time series.
Transcript of those detected in complete 30 year missing 20% data are ... · s e a wa t e r t e m p e ra t u re t...
Detectingmarineheatwaveswith sub-optimal dataRobert W. Schlegel 1,2,* @wiederweiter [email protected] Eric C. J. Oliver1 Alistair J. Hobday3 Albertus J. Smit2
1 Department of Oceanography, Dalhousie University, Halifax, Nova Scotia, Canada 2 Department of Biodiversity and Conservation Biology, University of the Western Cape, Bellville,South Africa 3 CSIRO Oceans and Atmosphere, Hobart, Tasmania, 7000, Australia
1 IntroductionThe Hobday et al. (2016) and Hobday et al. (2018)definitions are used here for marine heatwaves(MHWs) and their categories. These events areincreasing in duration and intensity around the globe(Oliver et al. 2018). Many coastal, oceanic, and polarregions with ecosystems/species that are vulnerableto these events (Smale et al. 2019) do not have optimalsea water temperature time series available for thedetection of MHWs. An optimal time series is defined here as meetingthese two primary requirements: 1) 30 years long, 2)no missing data. Any time series shorter than 30years long or missing any data is therefore sub-optimal. The purpose of this study was to determinehow confident we may be in our results when usingsub-optimal data.
1.1 Objectives
1. Find the sub-optimal data limits fordetecting MHWs
2. Quantify the e�ects of sub-optimal data onindividual MHWs
3. Determine if these e�ects are consistentaround the globe
2 MethodsReference time series from the heatwaveR package(Schlegel and Smit 2018) were re-sampled 100 times andmade increasingly sub-optimal by:
Removing one year of data; 37 – 10 yearsRandomly removing 1% of data; 0 – 50%Adding linear trend of 0.01°C/dec.; 0.00 – 0.50°C/dec.
The e�ect on an entire time series and one focus MHW(Figure 2.1) were measuredThe sub-optimal treatments were performed on globalNOAA OISST product and e�ects measured
Figure 2.1: Focus MHWs from the three reference time series: A) Western Australia, B)Northwest Atlantic, C) Mediterranean. Start/end dates shown with light-green marks.
3 ResultsMHWs detected with 10 years of data do not di�ersignificantly from those detected with 30 years of data(Figure 3.1) 20% missing data or less does not have an appreciablee�ect on the MHWs in a time series, but this can impactindividual MHWs by dividing them up into smaller events3.2) Individual MHWs in 10 year time series around the globeare on average 10% less intense (Figure 3.3) and 28%shorter than the same event in a 30 year time series
Figure 3.1: The e�ect of length on the MHWs detected within the 100 re-sampled timeseries. Note that lengths greater than 30 years also produce di�erent results.
Figure 3.2: The e�ect on the focus MHWs (Figure 2.1) given the three di�erent sub-optimaldata challenges.
Figure 3.3: Global map showing the trends in increasing or decreasing maximum intensityof MHWs as a time series is lengthened.
4 ConclusionsThe MHW detection algorithm provides robustresults that are insensitive to sub-optimal datachallenges. Missing data has a greater e�ect on MHWs than timeseries length. Linear interpolation is an e�ective fixfor missing data up to 40%. Time series length has a less predictable e�ect thanmissing data. The shifting up or down of the mean climatology isthe primary driver of change in MHWs. Optimal data should still be used when possible.
ReferencesHobday, Alistair J, Lisa V Alexander, Sarah E Perkins, Dan A Smale, Sandra C Straub, Eric CJ Oliver, JessicaA Benthuysen, et al. 2016. “A Hierarchical Approach to Defining Marine Heatwaves.” Progress inOceanography 141. Elsevier: 227–38.
Hobday, Alistair J, Eric CJ Oliver, Alex Sen Gupta, Jessica A Benthuysen, Michael T Burrows, Markus GDonat, Neil J Holbrook, et al. 2018. “Categorizing and Naming Marine Heatwaves.” Oceanography 31 (2).JSTOR: 162–73.
Oliver, Eric CJ, Markus G Donat, Michael T Burrows, Pippa J Moore, Dan A Smale, Lisa V Alexander, JessicaA Benthuysen, et al. 2018. “Longer and More Frequent Marine Heatwaves over the Past Century.” NatureCommunications 9 (1). Nature Publishing Group: 1324.
Schlegel, Robert W, and Albertus J Smit. 2018. “HeatwaveR: A Central Algorithm for the Detection ofHeatwaves and Cold-Spells.” The Journal of Open Source Software 3 (27): 821.
Smale, Dan A, Thomas Wernberg, Eric CJ Oliver, Mads Thomsen, Ben P Harvey, Sandra C Straub, MichaelT Burrows, et al. 2019. “Marine Heatwaves Threaten Global Biodiversity and the Provision of EcosystemServices.” Nature Climate Change 9. Nature Publishing Group: 306–12.
Marine heatwaves detected in time
series as short as 10 years and
missing 20% data are comparable to
those detected in complete 30 year
time series.
